mutex_atomic and CondVar

[Yamakaky]

The example at https://doc.rust-lang.org/std/sync/struct.Condvar.html issues a mutex_atomic warning, suggesting the replacement of Mutex::new(false) by an AtomicBool. My understanding is that we need a mutex to use CondVar.

[kaimast]

Still a problem in most recent nightly...

[gliderkite]

This is still reported as warning in Rust 1.45

[x4e]

Related to #4295.
I think perhaps this warning is slightly too easy to trigger. It would be far too complicated to statically determine if an AtomicBoolean is actually more appropriate than a Mutex. Perhaps this lint should be changed to an allow level to represent its high false positive rate?

[taiki-e]

Filed #8260 to downgrade this lint to nursery group that allowed by default.

[llogiq]

Is that the correct solution here?

In my experience I get a bunch of newbies who don't understand sync primitives triggering this lint and then there are a handful of people who actually know better. Making the lint allow by default will mean those in the know don't need to do anything while the others don't know anything to do.

I agree that the suggestion is suboptimal and should likely be removed in favor of better docs.

[taiki-e]

@llogiq
Personally, I agree with removing this lint, but it was proposed in the past and two maintainers effectively rejected it (#4295 (comment), #4295 (comment)).
However, no one has worked on fixing the bugs in this lint, and several people have written the wrong code due to this lint.
I don't have the energy to discuss this with the maintainers who are opposed to removing lint, so I suggested a downgrade as a compromise.

[llogiq]

Ok, so I'm going to accept the downgrade as a temporary measure until the suggestion and docs have been sorted out.